Architecture
data/models.yaml. Every label is auditable
against the model's sources.
Specs
- Architecture
- moe
- Total params
- not disclosed
- Active params
- 21B
- Experts
- 160 total · 6 active
- Context window
- not verified
- Attention
- mla
- Position encoding
- rope-yarn
- Post-training
- sft, rlhf
- OSI-approved
- no
- Data released
- no
- Training code
- not released
Benchmarks
Each score carries the date it was published; we never infer or interpolate missing scores.
Recommended use cases
- cost-efficient chat with code competence
- long-context retrieval
- function-calling agents
Available quantizations
GGUF llama.cpp's container; the common local format, k-quants from Q2 to Q8.
runs on llama.cpp, Ollama
Verified via the Hugging Face model tree ↗. Community quantizations change over time; the families shown are those with published weights at audit time.
Notable innovations
- · Chat + Coder track merger
- · Function calling and FIM in a single checkpoint
- · Safety score lifted to 82.6% (vs 74.4% for V2-0628)
Lineage
Combined the Chat and Coder tracks into one V2-architecture checkpoint, on the path to V3.
Derived from
DeepSeek-V2 Chat 2024-05-07Derivatives